Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 60, Issue 24, 2410007(2023)

A Low-Light Image Enhancement Method Combined with Generative Adversarial Networks in Nonsubsampled Shearlet Transform Domain

Wenling Shi, Yipeng Liao*, Zhimeng Xu, Xin Yan, and Kunhua Zhu
Author Affiliations
  • College of Physics and Information Engineering, Fuzhou University, Fuzhou 350108, Fujian, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (10)
    Equations (12)
    References (20)
    Tools
    Paper Information
    Topics
    Figures & Tables(10)
    NSST multiscale decomposition of low-light image

    Fig. 1. NSST multiscale decomposition of low-light image

    Download full size
    Structure of low-frequency image LF-EnlightenGAN enhanced model

    Fig. 2. Structure of low-frequency image LF-EnlightenGAN enhanced model

    Download full size
    Low-light image enhancement and denoising implementation flowchart

    Fig. 3. Low-light image enhancement and denoising implementation flowchart

    Download full size
    Effects of each step of low-light image enhancement and denoising

    Fig. 4. Effects of each step of low-light image enhancement and denoising

    Download full size
    Enhancement effect of each algorithm on the synthetic low-light images test set. (a) Low-light images; (b) ground truth; (c) LIME; (d) method in Ref. [6]; (e) ARD-GAN; (f) EnlightenGAN; (g) method in Ref. [12]; (h) proposed method

    Fig. 5. Enhancement effect of each algorithm on the synthetic low-light images test set. (a) Low-light images; (b) ground truth; (c) LIME; (d) method in Ref. [6]; (e) ARD-GAN; (f) EnlightenGAN; (g) method in Ref. [12]; (h) proposed method

    Download full size
    Denoising and edge detection effects of each algorithm on the synthetic low-light images test set. (a) Noise images; (b) LIME; (c) method in Ref. [6]; (d) RetinexNet; (e) ARD-GAN; (f) EnlightenGAN; (g) method in Ref. [12]; (h) proposed method

    Fig. 6. Denoising and edge detection effects of each algorithm on the synthetic low-light images test set. (a) Noise images; (b) LIME; (c) method in Ref. [6]; (d) RetinexNet; (e) ARD-GAN; (f) EnlightenGAN; (g) method in Ref. [12]; (h) proposed method

    Download full size
    Enhancement effect of each algorithm on the real low-light images test set. (a) Ground truth; (b) LIME; (c) method in Ref. [6]; (d) RetinexNet; (e) ARD-GAN; (f) EnlightenGAN; (g) method in Ref. [12]; (h) proposed method

    Fig. 7. Enhancement effect of each algorithm on the real low-light images test set. (a) Ground truth; (b) LIME; (c) method in Ref. [6]; (d) RetinexNet; (e) ARD-GAN; (f) EnlightenGAN; (g) method in Ref. [12]; (h) proposed method

    Download full size

    • Table 1. Enhanced performance statistics of each algorithm on the synthetic low-light images test set

      View table

      Table 1. Enhanced performance statistics of each algorithm on the synthetic low-light images test set

      Synthetic test setIndexLIMEMethod in Ref.［6］ARD-GANEnlightenGANMethod in Ref.［12］Proposed method
      Underwater imageSSIM0.94900.90000.87910.81870.93870.9507
      MSE0.01150.03760.02340.02990.01420.0049
      Normal light imageSSIM0.81970.89280.85800.84360.87320.8972
      MSE0.02140.02060.00630.00980.01700.0066
      Night imageSSIM0.79080.86270.86460.82810.86110.8650
      MSE0.01270.00620.00890.01370.00610.0055

    • Table 2. Denoising and edge detection performance statistics of each algorithm on the synthetic low-light images test set

      View table

      Table 2. Denoising and edge detection performance statistics of each algorithm on the synthetic low-light images test set

      Noise varianceIndexLIMEMethod in Ref.［6］RetinexNetARD-GANEnlightenGANMethod in Ref.［12］Proposed method
      10%PSNR /dB16.609916.822814.806422.067720.386317.612421.7001
      P /%78.870085.680084.800088.190090.920083.874291.1700
      20%PSNR /dB16.102416.711114.726621.834420.272917.524821.4184
      P /%73.370081.660081.450084.750087.840080.564089.7500
      30%PSNR /dB16.090916.128914.724220.044219.634617.058721.0829
      P /%67.770071.390078.210083.730082.810077.865188.1000

    • Table 3. Enhancement performance statistics of each algorithm on real low-light images

      View table

      Table 3. Enhancement performance statistics of each algorithm on real low-light images

      Quality assessment methodLIMEMethod in Ref.［6］RetinexNetARD-GANEnlightenGANMethod in Ref.［12］Proposed method
      Entropy7.69247.69137.54517.62697.68127.66517.7003
      BRISQUE15.252716.095315.993717.562215.255716.657411.4441
      ENIQA0.44190.44150.44210.43620.44520.44140.4413
      HyperIQA60.776752.226757.642659.290058.255156.784160.8533
    Tools

    Get Citation

    Copy Citation Text

    Wenling Shi, Yipeng Liao, Zhimeng Xu, Xin Yan, Kunhua Zhu. A Low-Light Image Enhancement Method Combined with Generative Adversarial Networks in Nonsubsampled Shearlet Transform Domain[J]. Laser & Optoelectronics Progress, 2023, 60(24): 2410007

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Apr. 7, 2023

    Accepted: Apr. 28, 2023

    Published Online: Nov. 27, 2023

    The Author Email: Liao Yipeng (fzu_lyp@163.com)

    DOI:10.3788/LOP231045

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology